“The only control or ‘guardrails’ AI needs is a STRONG AND SMART (High IQ!) PRESIDENT.”
Frontier AI labs want someone, anyone, to hit the brakes.
Healthcare wants to speed up.
Federal officials are reportedly considering paying autonomous AI doctors 60–80% of physician rates.
https://t.co/q61UGGk2o3
A Lancet Digital Health study tested 12 AI models on 33 clinical cases. Across 258,559 responses, models picked the advertised drug 47.6% of the time when an ad preceded the case, versus 35% without one. https://t.co/o40FtBaYd3
Healthcare delivery productivity fell about 1% from 1998 to 2023 while other service industries grew 56%, per McKinsey, which frames AI as the fix. In its own survey, 92% reported implementing or piloting generative AI, but only 38% report measurable ROI. https://t.co/L7yF61lZ7H
A 10-hospital study tested 7 LLMs on real medical records: DeepSeek-R1 and GPT-4o-reasoning tied for best diagnostic accuracy at 64%. Under simulated adversarial prompting, DeepSeek-R1 produced disinformation in 98% of cases, and even GPT-4o-reasoning did so 70% of the time. https://t.co/aEIgk2G0Kd
United Airlines is blocking unauthorized third-party AI agents from booking, citing risks of unintended purchases, exposed account data, and bypassed human checks like the bag-content confirmation at check-in. https://t.co/3z09dYuFYK
Medicaid work requirements under HR 1 begin Jan. 1, 2027, and the Commonwealth Fund estimates they could cut hospital margins up to 13.3% in expansion states. Health system leaders at a Becker's roundtable say that's pushing stricter ROI rules, hidden-cost scrutiny, and more vendor risk-sharing. https://t.co/NjxDYEU1ad
Pennsylvania's medical board filed what appears to be the first case under a medical practice act against an AI chatbot developer. The complaint alleges Character Technologies' "Emilie" persona claimed a psychiatry degree and a fake state license number across roughly 45,500 user interactions. https://t.co/f6nJcZU344
Patients are already using personal AI agents to try to book appointments, Emory's chief AI officer says. Health systems now have to choose: block those agents, let them in, or build their own automated scheduling tools instead. Emory saw traffic rise this summer but hasn't measured its volume yet. https://t.co/LUDRwpCpJd
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
A reported flaw in Claude Dispatch lets a browser extension controlling https://t.co/6ZzOozydGb run commands on a user's computer, or another machine on the same account, with no browser exploit needed. Anthropic confirmed the bug in May and reportedly still hasn't fixed it. https://t.co/gTXnQDXS3I
An NBC exit poll found 56% of voters called AI and data centers a major factor in their congressional vote. Democrats spent over $100 million tying GOP rivals to data centers and won in Ohio, Kansas, and Iowa, while Musk's $446 million in AI-aligned spending failed to save Texas's Ken Paxton. https://t.co/dfWbkPak2e
Medact says 29 of the 137 NHS trusts that England claims are "realising benefits" from Palantir's £330M data platform told FOI requests they weren't using it. One trust said FDP isn't live and has no plans to roll it out. Ministers decide on renewal in December. https://t.co/4UVtTEoY4h
Epic's AI costs now swing with token use, model choice, and feature volume. CIOs report per-feature charges ranging from $10 to $10,000. Patients linking MyChart to ChatGPT add usage hospitals didn't choose and still have to pay for. https://t.co/labB5dJz7W
FDA and CMS officials are working to head off a jurisdictional clash over clinical AI: states license clinicians and regulate medical practice, but AI that performs clinical tasks like writing prescriptions without human oversight can fall under FDA device authority. https://t.co/ks2nkYsMBu
Larry Ellison pitched Oracle Health AI modules—oncology, radiology, patient monitoring—as add-ons for Epic hospitals, claiming they can be switched on with no data migration and no new training. No health system has confirmed deploying them this way. https://t.co/3PgKrI4sLo
Survey of 146 Turkish hematologists: 87% used AI clinically in the past 3 months, but only 14% had formal training. Of those using generative AI for research, 81% hit fabricated references. 48.6% had entered patient data into public chatbots; just 4.8% had a written institutional AI policy. https://t.co/WAnEyag5Aq
AMA released a liability framework for clinical AI harms, tying responsibility to who can prevent it. Developers must prove safety and effectiveness; deployers answer for selection, oversight, and training; physicians are liable only for judgment they can meaningfully exercise. https://t.co/JWN2IqIRip
Northwell's Northern Market tried to replace three separate messaging tools with one unified platform. Nurses adopted it quickly, but physicians wanted only narrow texting features, and leaders paused the rollout after about a year of pushback. https://t.co/gpSkif1exp
Wolters Kluwer's 2026 survey finds 32% of nurses now use AI multiple times a day at work, self-reported, up from 16% a year ago. Top uses: summarizing literature (43%), patient education materials (41%), drug-interaction checks (37%). https://t.co/9gCZtJMsVm
CHOP now requires trainees to prove proficiency in tasks like note writing before they get access to AI tools, like scribes, that do them. Johns Hopkins and Iowa are applying similar gating to preserve independent clinical skills. https://t.co/AoQRl76Wy3